Oncogenehttps://doi.org/10.1038/s41388-019-0782-x

ARTICLE

Hypoxia-induced tumor exosomes promote M2-like macrophagepolarization of infiltrating myeloid cells and microRNA-mediatedmetabolic shift

Jung Eun Park1 ● Bamaprasad Dutta1 ● Shun Wilford Tse1 ● Nikhil Gupta1 ● Chee Fan Tan1 ● Jee Keem Low2●

Kheng Wei Yeoh3● Oi Lian Kon4

● James P. Tam1● Siu Kwan Sze 1

Received: 11 August 2018 / Revised: 8 February 2019 / Accepted: 4 March 2019© Springer Nature Limited 2019

AbstractDeveloping tumors rapidly outgrow their oxygen supply and are subject to hypoxia, which stimulates hypersecretion oftumor-derived exosomes that promote angiogenesis, metastasis, and immunosuppression, but the molecular mediators ofthese pathological effects remain poorly defined. Using quantitative proteomics, we identified that exosomes produced byhypoxic tumor cells are highly enriched in immunomodulatory proteins and chemokines including CSF-1, CCL2, FTH,FTL, and TGFβ. Modeling exosome effects on tumor-infiltrating immune cells, we observed a potent ability of thesehypoxia-induced vesicles to influence macrophage recruitment and promote M2-like polarization both in vitro and in vivo.In addition, hypoxic, but not normoxic, tumor exosomes enhanced oxidative phosphorylation in bone marrow-derivedmacrophages via transfer of let-7a miRNA, resulting in suppression of the insulin-Akt-mTOR signaling pathway. Together,these data demonstrate that hypoxia promotes tumor secretion of biomolecule-loaded exosomes that can modify theimmunometabolic profile of infiltrating monocyte-macrophages to better evade host immunity and enhance tumorprogression.

Introduction

Low oxygen availability or ‘hypoxia’ stress in developingtumors stimulates secretion of small, membrane-boundvesicles known as exosomes that can promote

angiogenesis, metastasis, and suppression of host immunityto drive disease progression [1, 2]. While exosomes canmediate steady-state communication between cells and themicroenvironment in healthy tissues [3], their ability todeliver proteins, DNA, and RNAs to host leukocytes hasalso been strongly implicated in several key pathologicalprocesses [4, 5]. It is now clear that exosomal transfer ofbioactive molecules allows tumors to influence the behaviorof infiltrating immune cells, but the molecular mediatorsand mechanisms by which the exosomal cargo supportsdisease progression remain poorly defined.

An essential role of exosomes in tumorigenesis hasalready been demonstrated in breast cancer, gastric cancer,and melanoma [6–8]. However, the biomolecule cargo ofthese exosomes is highly diverse and can exert variableeffects on tumor angiogenesis, cell migration, and primingof the pre-metastatic niche via interactions with healthystromal tissues [6, 9, 10]. Some researchers have reportedthat tumor exosomes are immunogenic and elicit specifichost responses that might influence disease outcome. Forexample, tumor antigen-loaded vesicles derived from den-dritic cells and macrophages have been shown to exhibit T-cell stimulatory activity mediated by transfer of MHC I/II,

* Jung Eun Parkjepark@ntu.edu.sg

* Siu Kwan Szesksze@ntu.edu.sg

1 School of Biological Sciences, Nanyang Technological University,60 Nanyang Drive, Singapore 637551, Singapore

2 Department of Surgery, Tan Tock Seng Hospital, 11 Jalan TanTock Seng, Singapore 308433, Singapore

3 National Cancer Centre Singapore, Department of RadiationOncology, 11 Hospital Drive, Singapore 169610, Singapore

4 National Cancer Centre Singapore, Division of Medical Sciences,11 Hospital Drive, Singapore 169610, Singapore

Supplementary information The online version of this article (https://doi.org/10.1038/s41388-019-0782-x) contains supplementarymaterial, which is available to authorized users.

1234

5678

90();,:

1234567890();,:

CD54, CD80, and CD86 [11, 12]. Similarly, exosomesisolated from primary human body fluids have beenobserved to activate monocytic cells via toll-like receptorsignaling [13]. However, the majority of prior studies havereported that tumor-derived exosomes are typically immu-nosuppressive [14]. In leukemia, tumor exosomes exhibitmembrane-bound TGF-β1 and disrupt NK cell activity tomediate immune evasion [15, 16], while in gastric cancerand melanoma, exosomes display variable lymph nodelocalization and disease-enhancing effects depending ontheir source [7, 10, 17]. Taken together, these data indicatethat tumor exosomes can exert potent effects on hostimmunity and are likely to represent key determinants of theclinical course in human cancer.

Numerous studies have investigated the role of tumor-derived exosomes in cancer progression, but little is knownabout how the tumor microenvironment alters the content ofexosome vehicle and the consequence of these alterations inthe context of immune response. Since microenvironmentalstress is already known to modulate exosomal sortingmechanisms in healthy cells and tissues [18, 19], we hypo-thesized that low oxygen availability also modifies themolecular composition of tumor exosomes. Indeed, given thathypoxia is recognized as a key regulator of tumor progression,it is likely that exosomes secreted by hypoxic cancer cellsexhibit immunomodulatory effects and induce micro-environmental changes to support tumor development [1, 20].In the current report, we identify a potent ability of hypoxictumor exosomes to promote M2-like polarization of infiltrat-ing macrophages driven by microRNA (miRNA)-mediatedinduction of an oxidative phosphorylation (OXPHOS) meta-bolic profile that favors tumor development.

Results

Hypoxia enhances tumor secretion of exosomesenriched in immunosuppressive proteins

To explore the immunological effects of tumor exosomessecreted under hypoxic conditions, we exposed mousemelanoma B16-F0 cells to normoxia (21% O2) or hypoxia(<0.5% O2) for 24 h in a hollow fiber cell culture systemthat generates a 3D environment to mimic conditionsin vivo [21]. Exosomes isolated from the cell culturesupernatants were then subjected to tandem mass tags(TMT)-based quantitative proteomic analysis as describedin the Materials and Methods and shown in Fig. 1a [22].Nanoparticle tracking analysis of the isolated exosomesrevealed an average diameter of 136 ± 42 nm, consistentwith the 30−150 nm size range expected for these vesicles(Fig. 1b, Supplementary Videos 1 and 2), while westernblot analysis confirmed the presence of characteristic

exosomal markers Alix, CD63, and TSG101 (Fig. 1c).Exosome size distribution did not differ between normoxiaand hypoxia conditions, but total exosomal proteins percells were 3–4-fold higher under low-oxygen conditions infour cancer cell lines (Fig. 1d). Similar data were alsoobtained when assessing the impact of hypoxia on exosomesecretion by nanoparticle tracking system in four cancer celllines (Supplementary Fig. S1), thus indicating that oxygenrestriction exerts potent effects on exosome loading acrossmultiple tumor types. We next proceeded to identify andquantitate the exosomal proteins obtained using Mascotdatabase search software (specific search criteria aredescribed in the Materials and Methods). A total of 3710proteins were identified based on at least two unique trypticpeptides and protein score ≥65 (p < 0.05; SupplementaryTables S1 and S2). Enrichment ratio was calculated asprotein abundance in hypoxic exosomes relative to nor-moxic exosomes (log2 fold change ≤−1 and ≥1 wereassigned as cut-off values based on volcano plot analysis;Fig. 1e, Supplementary Fig. S2). Using this approach, weobserved that a total of 990 proteins (26.7% of total exo-somal proteins) were at least 2-fold enriched in hypoxicexosomes, whereas only 220 proteins (5.9% of total) weremore abundant in normoxic exosomes.

Microenvironmental hypoxia was associated with tumorexosome enrichment in several proteins already implicatedin tumor–tumor or tumor–stroma interactions includingchemokines, growth factors, pro-tumorigenic molecules,and various soluble inhibitory factors (Fig. 2a–d and Sup-plementary Table S3). Key components of these vesiclesincluded the chemokines/chemoattractants macrophagecolony-stimulating factor 1 (CSF-1), monocyte chemoat-tractant protein-1/C–C motif chemokine 2 (MCP-1/CCL2),endothelial monocyte-activating polypeptide 2 (EMAP2/AIMP1), and leukotriene A-4 hydrolase (LTA4H). We alsoobserved marked exosomal enrichment in immunosup-pressive mediators including transforming growth factorbeta 1 (TGFβ1), TGFβ2, TGFβ3, macrophage migrationinhibitory factor (MIF), and ferritin heavy/light chain (FTH,FTL). In addition, hypoxic tumor exosomes were furtherloaded with proteins that might contribute to tumor pro-gression/metastasis and miRNA processing includingmatrix metalloproteinases 2 and 19, annexin A4 (ANXA4)and ANXA6, procollagen-lysine,2-oxoglutarate 5-dioxygenase 1 (PLOD1), PLOD2, argonaute 1 (AGO1),AGO3, and AGO4. In Table 1, we have summarized theroles of these enriched proteins in hypoxic exosomes. Keyproteins enriched in hypoxic exosomes were further con-firmed by western blot analysis in various cancer cell lines.As shown in Fig. 2e and Supplementary Fig. S2, theexpression levels of CSF-1, FTH, or FTL were increased inhypoxic exosomes, which were isolated from B16-F0,A375, A431, and A549 lung adenocarcinoma cells.

J. E. Park et al.

Previous studies have suggested that tumor-derived exo-somes are loaded with immunosuppressive or pro-tumorigenic molecules [11, 14], but the exact role of exo-somes released into the tumor microenvironment is stillcontroversial. To further investigate the relation betweentumor hypoxia and composition of the exosomal cargo, weanalyzed protein co-expression level with the master reg-ulator of hypoxia HIF1α and key exosomal proteins [23]. Inorder to do this, we used the computational gene co-expression search engine SEEK [24] to access publiclyavailable microarray datasets, which revealed a close asso-ciation of HIF1α with expression of target genes includingCCL2, CSF1, TGFβ2, TGFβ3, FTH1, PLOD2, LGALS3,and MIF (Fig. 2f), across multiple tumor types includingglioblastoma, lung, melanoma, and pancreatic cancers(Supplementary Table S4). Of these hypoxia/HIF1α-responsive genes, CSF-1 and CCL2 are known to be keymediators of monocyte/macrophage recruitment into tumorsand contribute to the establishment of a pro-tumorigenicmicroenvironment [25, 26], while TGFβ and FTH1 potentlysuppress host immunity via induction of regulatory T-cells(Tregs) [27, 28]. Coordinated dispersal of these proteins viaexosome release during tumor hypoxia is therefore likely topromote immune evasion and disease progression.

Hypoxic tumor exosomes promote M2-likepolarization of infiltrating monocyte-macrophages

Our proteomics data revealed that hypoxic exosomes werehighly enriched in potent chemoattractants for monocyte/macrophages (CSF-1, CCL2, and EMAP2), which canundergo differentiation into tumor-associated macrophages(TAMs) in the tumor microenvironment and promote cancerprogression [29, 30]. We therefore assessed whether hypoxictumor exosomes are capable of altering leukocyte recruitmentby assessing their impact on cell motility in a standard Boy-den chamber assay [31]. Macrophage-like RAW 264.7 cellsexhibited 2- to 5-fold greater chemotactic responses toexosome-containing media than to PBS-only control medium(Fig. 3a) and cell migration was significantly enhanced whenthe bottom chamber contained exosomes from hypoxictumors (4.3-fold, p < 0.0001). These data suggested thathypoxic tumors exhibit greater capacity to recruit monocyte/macrophages via exosome release into the local micro-environment. In order to assess this possibility in vivo, wenext mixed B16-F0 tumor cells (1 × 106 cells) with normoxicexosomes, hypoxic exosomes, or matrigel-only control priorto subcutaneous (s.c.) injection to the backs of C57BL/6 mice(n= 6 animals per group). On day 14 after s.c. injection of

Fig. 1 Schematic diagram of hollow fiber cell culture system andproteomic workflow. a Schematic diagram of hollow fiber cell culturesystem and workflow of exosome extraction for proteomic analysis. bNanoparticle tracking analysis of exosomes isolated from the extra-capillary space (ECS) indicating mean diameter of 136 ± 42 nm. cWestern blot analysis confirmed the presence of exosomal markerproteins, Alix, CD63, and TSG101. Ponceau staining was used as an

equal loading control. d Comparison of total exosomal proteinssecreted by normoxic cancer cells, B16-F0, A375, A431, A549, ortheir hypoxic counterparts. Data are displayed as mean ± SEM, ***p <0.0001. e Histogram shows the distribution of fold-change in exosomalproteins comparing hypoxia with normoxia (Log2 [Hx/Nx] ≤−1 or ≥1are indicated). NxExo normoxic exosomes, HxExo hypoxic exosomes

Hypoxia-induced tumor exosomes promote M2-like macrophage polarization of infiltrating myeloid cells. . .

B16-F0 cells, the mice were sacrificed and the tumor masswas isolated and dissociated for analysis by flow cytometry toevaluate M1- or M2-like macrophage recruitment/differ-entiation within the tumor area. As shown in Fig. 3b, tumorinfiltration of F4/80+CD206high M2-like macrophage [32] wassignificantly increased by co-injection of hypoxic exosomes(fold change 3.93, p= 0.0036) compared to matrigel-onlycontrol, whereas only trace numbers of dendritic cells(CD11chi, lower right quadrant) or immunosuppressive Tregs(CD4+CD25+) were detected in the tumor region (Supple-mentary Fig. S3).

Since macrophages derived from myeloid precursor cellscan exhibit pro-tumor or immunosuppressive activity uponCSF-1 or CCL2-mediated differentiation to M2-like macro-phage or TAM in tumor local area, we further characterizedthe exosome-induced differentiation of myeloid cells usingM2-like macrophage specific marker, CD206 [30]. Exosome-treated bone marrow-derived macrophages (BMMs) showed ashift of cell population to F4/80+CD206high population com-pared to control BMMs. Especially, hypoxic exosome pro-moted the expansion of F4/80+CD206high population by 1.74-fold compared to PBS control without IL-4 treatment,

Fig. 2 Summary of exosomalproteins associated withimmunosuppression and/ortumor progression. a Schematicoverview of identified proteinsthat were enriched in exosomesreleased by hypoxic B16melanoma cells. b–d Hypoxia-enriched exosomal proteins wereclassified into three maincategories; b chemokines andgrowth factors, c solubleinhibitory proteins, and d pro-tumorigenic proteins andmiRNA-processing proteins.Protein fold-changes are shownin log-scale. e Western blotanalysis of key exosomalproteins, CSF-1, FTH, and FTL.f Analysis of HIF-1α co-expression with target genesCCL2, CSF1, TGFβ2, TGFβ3,FTH1, PLOD2, LGALS2,LGALS3, and MIF (as assessedusing SEEK software)

J. E. Park et al.

indicating a key role of tumor-derived exosome on M2-likemacrophages induction (Fig. 3c). In addition, BMMs pre-sented high F4/80+CD135high phenotype, a hematopoieticstem cell marker, upon exosome treatment [33], implying thepotential role of exosomes in stem cell development (Fig. 3d).IL-4 (20 ng/ml) was treated for M2-type macrophage polar-ization (Supplementary Fig. S4). Hypoxic exosome-mediatedM2-like macrophage expansion was in line with BMMs cellgrowth. As shown in Fig. 3g, hypoxic exosome-treatedBMMs showed 1.7-fold increase in cell number compared toM-CSF-only treated cells (p= 0.0027).

Given that macrophages can exhibit either anti-tumor(M1) or pro-tumor polarization (M2) upon cytokine-mediated activation of STAT1 or STAT6 signaling path-way, respectively [34, 35], we next sought to determinewhether tumor exosomes can mediate this process bywestern blot analysis (Fig. 3e). While STAT1 phosphor-ylation was barely detected in either PBS- or exosome-treated BMMs, STAT6 phosphorylation was substantiallyincreased in exosome-treated BMMs, consistent with anability to promote M2-like polarization of tumor-infiltratingmacrophages. Accordingly, expression of M2-like markergenes assessed by quantitative real-time PCR (RT-qPCR)showed that exposure to hypoxic exosomes for 48 h resulted

in time-dependent upregulation of Arg1 (arginase 1), Ym1(chitinase 3-like 3), and Fizz1 (found in inflammatory zone1) (Fig. 3f) [36].

To further explore the effects of cancer exosomes on M2-like macrophage polarization, exosomes were isolated fromvarious cancer cell lines including A375, A431, and A549,and treated to THP-1 monocytic cells to test the expressionof M2 markers. For this, THP-1 cells were treated with20 ng/ml of PMA (Phorbol 12-myristate 13-acetate) for 24 hfollowed by resting in RPMI control media for 24 h todifferentiate into M0-like macrophages. Once differentiated,M0-type THP-1 cells can be polarized to either M1- or M2-like macrophage response to different inducers [37].Afterward, 5 µg/ml of cancer exosomes were treated to M0-type THP-1 cells for 48 h and applied to either FACSanalysis or RT-qPCR analysis for M2-type markers such asthe scavenging receptor CD163 or chemokine CCL13 [38,39]. With PMA treatment, a clear increase in the level ofCD163 was found in THP-1 cells, and hypoxic exosometreatment increased the CD163-positive population up to55% (Supplementary Fig. 6A). In addition, the expressionlevel of CCL13 was upregulated in hypoxic exosome-treated THP-1 cells (Supplementary Fig. 6B), which indi-cated that exosomes derived from different hypoxia cancer

Table 1 Role of the keyexosomal proteins secreted fromhypoxic cancer cells

Protein Role in tumor/tumor microenvironment Log2(Hx/Nx)ratio

Refs.

CSF1 Monocyte/macrophage recruitmenMacrophage (TAM) differentiationPoor prognosis

2.663 [54–56]

CCL2 Monocyte/macrophage recruitmentCancer progression and metastasis

1.234 [57, 56]

EMAP2/AIMP1 Macrophage recruitmentTAM accumulation

0.968 [58, 55]

LGALS3 M2 macrophage infiltration and angiogenesis in tumors 1.614 [59]

FTH Immunosuppression, activation of regulatory T cell 2.545 [60, 61]

FTL Immunosuppression, Tumor development, and prognosticbiomarker

2.471 [62–64]

TGFβ 1TGFβ 2TGFβ 3

Immune evasion, cancer stemness, and metastasis 1.0801.3421.312

[65–67]

LTA4H Chronic inflammation-associated carcinogenesis 1.104 [68, 69]

MIF Immunosuppression, tumor development 0.788 [70, 71]

HDGF Cancer cell invasion 1.905 [72]

MMP2MMP19

Cancer progression 1.81.096

[73]

PLOD1 Tumorigenesis/cancer biomarker 1.778 [74]

PLOD2 Hypoxia-induced tumor metastasis 1.693 [75]

ANXA4 Tumor progression, invasion, metastasis, and drug resistance 1.412 [76]

AGO4AGO3AGO1

Processing of small RNA such as miRNA, siRNA for genesilencing

1.2641.0681.104

[77, 78]

ANXA6 Dual functions, acting as a tumor suppressor or promoter 1.033 [79, 80]

Hypoxia-induced tumor exosomes promote M2-like macrophage polarization of infiltrating myeloid cells. . .

cell lines indeed contribute to M2-type macrophage polar-ization. Collectively, hypoxic tumor exosomes facilitatemonocyte/macrophage recruitment both in vitro and in vivo,

and drive protumoral M2-like macrophage polarization ofthe infiltrating cells to potentially enhance cancerdevelopment.

J. E. Park et al.

Hypoxic exosomes enhance mitochondrial OXPHOSin macrophages

M2-like macrophages have previously been reported toemploy mitochondrial OXPHOS to sustain immune func-tion, whereas M1-like macrophages and cancer cells useglycolysis to generate ATP [34, 40]. Metabolic reprogram-ming assists macrophages in adapting to their local micro-environment, thus shaping their activation state/effectorfunctions and potentially impacting on disease progression.We therefore assessed whether exosome-mediated polar-ization of macrophages toward an M2-like profile wasassociated with changes in metabolic activity. We examinedthe metabolic shift in exosome-treated BMMs by measuringATP-linked mitochondrial oxygen consumption rate (OCR)in response to oligomycin (OM), carbonyl cyanide-4-(tri-fluoromethoxy)-phenylhydrazone (FCCP), and rotenone(ROT)+ antimycin A (AA). M2-like macrophages polarizedby hypoxic exosomes exhibited enhanced OXPHOS activityat both basal respiration and ATP production level relative tocells treated with normoxic exosomes or PBS-only control(Fig. 4a, b), indicating that BMMs exposed to hypoxicexosomes undergo a metabolic shift toward TCA cycleactivation and OXPHOS, which are archetypal character-istics of M2-like polarization [34]. We next assessed sig-naling pathways that might drive a metabolic shift fromglycolysis to OXPHOS such as AKT-mTOR signalingpathway. Western blot analysis showed that hypoxic

exosome-treated BMMs displayed reduction of AKT andmTOR phosphorylation while total AKT and mTOR werenot affected and phosphorylation of p4E-BP and pS6K,direct substrates of mTOR activation, was subsequentlyreduced (Fig. 4c), together with increased expression levelsof mTOR negative regulator REDD1 (Fig. 4d).

Exosomal let-7a miRNA regulates the mTORsignaling pathway in macrophages

Having observed that tumor-derived exosomes enhancedmitochondrial OXPHOS and suppressed mTOR activationin BMMs, we next sought to identify exosomal factors thatmight influence mTOR activity and/or glucose metabolism.Recently, miRNAs have emerged as key regulators of thesesignaling pathways to regulate cell metabolic process [41,42], so we used RT-qPCR to assess the expression level ofcandidate miRNAs including let-7a and miR-21a in tumor-derived exosomes. Under hypoxic conditions, total let-7amiRNA expression in tumor cells was decreased to just~30% of that detected in the normoxic control, whereasexosomal let-7a content was increased ~25-fold (Fig. 5a).These data indicate that let-7a miRNA is substantiallydecreased in hypoxic tumors via export from the cellsmediated by exosome release. Exosomal miR-21a level wasslightly increased in hypoxic exosomes (Supplementary Fig.S7). We therefore sought to explore the expression level oflet-7a target genes in BMMs by RT-qPCR after exosometreatment. miRNA let-7a target genes, which involved ininsulin signaling pathway, were predicted using TargetScansoftware (Fig. 5b) [43]. Using this approach, we observedthat exposure to tumor exosomes significantly decreased theexpression levels of let-7a target genes such as IRS-1, IRS-2,INSR, and IGF1R in BMMs (Fig. 5c). Consistent with thesefindings, when let-7a miRNA mimics was transfected intomacrophage-like RAW 264.7 cells, we observed a sig-nificant decrease in the expression levels of INS-1 andIGF1R, supporting that transfer of let-7a miRNA into targetmacrophages can mediate potent suppression of insulinsignaling molecules (Fig. 5d). Since metabolic shift ofimmune cells affects the immune cell fate and function, wefurther analyzed key functional molecules in BMMs afterexosome treatment. As shown in Fig. 6a, hypoxic exosome-treated BMMs displayed 30- to 80-fold higher expression ofthe TAM-associated genes COX-2, PGES-1, and IL-6, whichhave established roles in host immunosuppression and tumorgrowth. In contrast, normoxic tumor exosomes exerted onlymodest effects on macrophage expression of these genes(Fig. 6a). Interestingly, tumor-derived exosomes showedprofound effect on BMMs but not on bone marrow-deriveddendritic cells (BMDCs). Under the same treatment condi-tion, BMDCs showed around 3–10-fold increase of COX-2,PGES-1, IL-6 gene expression compared to control, but

Fig. 3 Tumor-derived exosomes induce M2-like polarization of bonemarrow-derived macrophages. a Macrophage-like RAW 264.7 cellswere added to the Matrigel-coated upper wells of a Boyden chamber forassessment of chemotaxis toward normoxic or hypoxic exosomes addedto the lower chamber. The cells were allowed to migrate for 24 h at 37 °Cbefore staining with crystal violet and quantification. Data representmean ± SEM of triplicate experiments, ***p < 0.0001. b Immune cellinfiltration analysis in tumor xenograft model. B16-F0 cells were pre-mixed with 50 µg of hypoxic or normoxic exosomes (or PBS-onlycontrol) prior to s.c. injection into C57BL/6 mice (n= 6 animals pergroup). On day 14, the resultant tumors were excised/dissociated and thephenotype/frequency of infiltrating Mφ was assessed by flow-cytometry.Representative dot-plots show F4/80+CD206high phenotype in B16tumors co-injected with PBS control, normoxic, or hypoxic exosomes. c–f Exosome-mediated M2-like polarization of BMMs. BMMs were co-cultured with 5 μg/ml of normoxic or hypoxic exosomes for 48 h andsubsequently used for FACS analysis. PBS was used as a control group.c Data showed relative expression of F4/80+CD206high on BMMs. dData showed the relative expression of F4/80+CD135high on BMMs. eWestern blot analysis of STAT1, STAT6, p-STAT1 Y701, and p-STAT6Y641 in exosome-treated macrophages. f RT-qPCR analysis of M2-likemarker genes Arg1, Fizz1, and Ym1. Data represent mean ± SEM oftriplicate experiments, ***p < 0.0001. g Hypoxic tumor exosomesenhance the cell growth of BMMs. Bone marrow cells were culturedwith M-CSF for 6 days to differentiate into BMMs and then treated with5 µg/ml of NxExo or HxExo for 2 days. Trypan Blue dye exclusion testwas used for viable cell counting. Hypoxic exosome treatment promotedcell growth of BMMs; **p= 0.0027

Hypoxia-induced tumor exosomes promote M2-like macrophage polarization of infiltrating myeloid cells. . .

extremely lower than BMMs’ gene expression changes. Inaddition, BMMs showed higher response to hypoxic con-dition when being co-cultured with B16-F0 cells by pre-senting more than 200-fold increase of COX-2 and IL-6 geneexpression, while BMMs alone induce only ~4-fold changesunder hypoxic culture (Fig. 6c). It implies that tumor-infiltrating immune cells, specifically macrophages, stronglyreact with secretory molecules derived from hypoxic tumorcells rather than hypoxic condition. Since an exosome hasbeen demonstrated as a stable delivery vehicle of biomole-cules, the exosome might take a key role in immune cellmodulation in the tumor microenvironment. This specificM2-like polarization of exosome-treated BMMs, combined

with our earlier observation of enhanced OXPHOS in thesecells, strongly suggested that macrophages generated via thisprocess might favor cancer progression. In order to test this,we cultured BMMs for 48 h in the presence of normoxic orhypoxic exosomes and then tested whether soluble factorsreleased into the conditioned medium (CM) were capable ofaltering the tumor growth. When assessed by 3-(4,5-dime-thylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT)assay, we observed that BMM-derived CM significantlyincreased B16-F0 tumor cell proliferation and viability, butgrowth rates were enhanced 3- to 4-fold more after transferof CM generated by hypoxic exosome-treated macrophages(Fig. 6d).

Fig. 4 Exosome-mediated enhancement of mitochondrial oxidativephosphorylation. a Oxygen consumption rate (OCR) of BMMs treatedwith NxExo, HxExo, or PBS control using mitochondrial stress test.OCR was tested in biological triplicates. b Basal respiration rate andATP production were presented in biological triplicates. Paired T-testbetween PBS- and HxExo-treated BMMs was performed after

combining three biological replicates. c Western bot analysis ofphospho-Akt and phospho-mTOR as well as substrates p70S6K and4E-BP1. d RT-qPCR analysis of REDD-1 expression level. Datarepresent the mean ± SEM of triplicate experiments, ***p < 0.0001.NxExo normoxic exosomes, HxExo hypoxic exosomes

J. E. Park et al.

Discussion

Exosomes are now widely recognized as critical mediatorsof tumor growth, angiogenesis, and metastasis, and haverecently become a major focus of efforts to develop newcancer therapeutics. However, it remains unclear how exo-some biology is modified by dynamic changes in the tumormicroenvironment. Since tissue hypoxia is known to be akey regulator of tumor development [1, 44], we assessedhow oxygen depletion influences tumor exosome release,protein composition, and effects on immune responses bothin vitro and in vivo. Using a comprehensive proteomicapproach, we observed that tumor exosomes are enriched inchemokines and growth factors that mediate monocyte/macrophage recruitment and host immunosuppressionincluding CSF-1, CCL2, EMAP2, TGFβ, FTH, and FTL[25–28, 45, 46], and hypoxia substantially enhanced exo-somal levels of these proteins 4–6-fold higher comparedwith vesicles generated under normal oxygen conditions(Fig. 2). These data suggest that exosomes released byhypoxic tumors may exert more potent effects on hostimmunity and alter the clinical course of human cancers.

In this study, the immunological effects of tumor exo-somes were further investigated using bone marrow-derivedmyeloid cells, an origin of circulating and tissue-residentmonocyte/macrophages [47]. We observed increased mac-rophage chemotaxis toward exosomes produced by hypoxictumors relative to those released under standard cultureconditions. In addition, hypoxic exosomes increased M2-like polarization of bone marrow-derived myeloid cells (F4/80+CD206high positive cells) without cytokine supple-mentation (Fig. 3c) and promoted change in macrophageimmunometabolic profile to enhanced OXPHOS activity(Fig. 4) [48, 49]. These findings are consistent with earlierreports that M2-like macrophages can promote tumor pro-liferation and enhance the angiogenic potential of endo-thelial cells by favoring OXPHOS [50]. In parallel withthese immunometabolic effects, hypoxic exosomes sig-nificantly increased the expression of M2-associated genesimplicated in supporting cancer cell growth (Arg-1, IL-6,COX2, and PGES), and CM obtained from these cells wascapable of transferring an enhanced growth phenotype tofresh tumor cells that had not been directly exposed toexosomes.

Fig. 5 Exosome-mediated transfer of miRNA let-7a to BMMs. a RT-qPCR analysis of let-7a miRNA levels in B16-F0 tumor cells andderivative exosomes. Data represent the mean ± SEM of triplicateexperiments; ###p= 0.0002, ***p < 0.0001. b Proposed model for therole of exosomal let-7a miRNA in INS-AKT-mTOR pathway. Can-didate let-7a miRNA target genes as predicted by TargetScan; insulin-like growth factor 1 receptor (IGF1R), insulin receptor (INSR), insulinreceptor substrate-1 (IRS-1), IRS-2, insulin-like growth factor-1 (IGF-

1), insulin-like growth factor 2 mRNA-binding protein 1 (IGF2BP1),IGF2BP2, and IGF2BP3. c RT-qPCR analysis of let-7 miRNA targetgene expression in exosome-treated BMMs. Data represent the mean± SEM of triplicate experiments; #p= 0.0384, *p= 0.0234. d RT-qPCR gene expression analysis of RAW264.7 macrophages aftertransfection with 50 nM of let-7a mimics. Data represent the mean ±SEM of triplicate; **p= 0.0186, ##p= 0.0036

Hypoxia-induced tumor exosomes promote M2-like macrophage polarization of infiltrating myeloid cells. . .

Our findings of exosome-mediated M2-like macrophagepolarization were supported by hypoxic tumor-derivedexosomes from various cancer cell lines including humanmelanoma, skin, and lung cancer cells. Tumor-derivedexosomes, especially hypoxic exosomes, induced M2-likephenotypic changes from M0-like THP-1 cells, assessed byCD163+ marker and the expression of CCL13 (Supple-mentary Fig. 6), indicating the global role of tumor-derivedexosomes in immune systems. Interestingly, tumor exo-somes exhibited a potential role of hematopoietic stem celldifferentiation as it increased F4/80+CD135high positive cellpopulation in exosome-treated BMMs (Fig. 3d).

Exosome-enhanced OXPHOS activity in BMMs wasassociated with suppression of the mTOR pathway, sug-gesting a role of exosomes via inhibition of AKT-mTORsignaling pathway. In the present study, we identified let-7amiRNA, a known epigenetic tumor suppressor [51], as acandidate suppressor of insulin-mediated mTOR signalingpathway. Our data indicate that microenvironmentalhypoxia was instead associated with decreased let-7amiRNA levels in cancer cells and a >20-fold increase inreleased exosomes, suggesting that this may represent a

mechanism of expelling suppressive molecules from thetumor body. In turn, exosomal transfer of let-7a miRNAenhanced OXPHOS activity and M2-like polarization ofinfiltrating macrophages via downregulation of insulin-AKT-mTOR signaling pathway. While let-7a miRNA wasenriched in hypoxic exosomes, the suppressive effects ofexosomes on insulin signaling-related genes were sig-nificant in both normoxic and hypoxic exosome-treatedcells. Therefore, further characterization of let-7a miRNAtargets as well as hypoxic exosome-specific factors isnecessary. These combined effects significantly increasedtumor cell proliferation and survival in our assays in vitro,and are therefore also likely to exert major effects on cancerprogression in human patients in vivo.

M2-like macrophage polarization and change in immu-nometabolic profile are now recognized as critical compo-nents of tumor progression and represent appealing targetsfor novel immunotherapies. In the current report, we pro-vide a mechanistic link between the tumor microenviron-ment and M2-like polarization of infiltrating macrophagesvia exosomal transfer of immune mediators and suppressivemiRNAs. Our proteomics data revealed that hypoxia exerts

Fig. 6 Exosome-treated BMMs increased cancer cell growth. a RT-qPCR analysis of M2-associated genes, COX-2, PGES, and IL-6, inexosome-treated BMMs. Data represent the mean ± SEM of triplicateexperiments; ***p < 0.0001. b RT-qPCR analysis of COX-2, PGES,and IL-6 in exosome-treated BMDCs. Data represent the mean ± SEMof triplicate experiments; ***p < 0.0001. c RT-qPCR analysis of COX-2 and IL-6 in BMM-B16-F0 co-culture system. Data represent the

mean ± SEM of triplicate experiments; ***p < 0.0001. d B16-F0 tumorcell viability as assessed by MTT assay after 48 h culture with CMgenerated by BMMs exposed to normoxic exosomes (NxExo) orhypoxic exosomes (Hx-Exo) for the preceding 48 h (or PBS-only/normal medium control). Data represent the mean ± SEM of triplicateexperiments

J. E. Park et al.

profound effects on the composition of exosomal cargo,with many proteins displaying >2-fold change in abundanceunder these conditions. Since tumor hypoxia has long beenassociated with poor prognosis/therapy resistance, andexosome levels in serum are substantially increased incancer, it is likely that hypoxia effects on exosome loadingand release will significantly impact on disease progressionin human patients. These findings also suggest that tumorexosome release into the circulatory system may represent asource of potential biomarkers for predicting clinical courseand informing treatment strategies. Further studies will nowbe required to better define the role of hypoxic tumorexosomes in promoting therapy resistance in humanpatients, so that more effective immunotherapeuticapproaches can be developed in future.

Materials and methods

Hollow fiber cell culture

Mouse melanoma B16-F0 cells and Raw 264.7 macro-phage cells and human melanoma A375 cells, squamousskin carcinoma A431 cells, lung adenocarcinoma A549cells, and monocytic THP-1 cells were purchased fromAmerican Type Culture Collection (ATCC, Manassas,VA, USA) and maintained in DMEM-high glucose mediaor RPMI 1640 (HyClone Laboratories, Logan, UT, USA)supplemented with 10% fetal bovine serum at 37 °C with5% CO2. All cell lines were tested free of mycoplasmacontamination [52].

For exosome isolation, we used a hollow fiber culture(HFC) system to generate an in vivo-like 3D environment.All hollow fiber modules including Fibercell cartridges(#C2011) were purchased from Fibercell Systems Inc.(Frederick, MD, USA). Briefly, B16-F0, A375, A431, orA549 cells were inoculated into the extra-capillary space ofeach cartridge and allowed to attach to the fiber surface over24 h. Cancer cells were maintained in DMEM supple-mented with 5% Chemically Defined Medium for HighDensity cell culture (CDM-HD; Fibercell). CDM-HD, achemically defined protein-free serum replacement, wasused to avoid contamination of FBS-derived exosomes.After cell inoculation, the media reservoir was refreshedupon reaching 50% of initial glucose concentration, afterwhich the cultures were subjected to either normoxia (Nx,21% O2) or hypoxia (Hx, 0.5% > O2) for 24 h duration.Presence of oxygen was checked by Mitsubishi RT Anaero-Indicator (Thermo Fisher Scientific).

Exosome extraction, purification, and quantification

CM was collected from the hollow fiber cartridge’s extra-cellular space after 24 h Nx or Hx culture, before beingcentrifuged at 1200 × g, 25 °C for 30 min to remove celldebris. The supernatants were then subjected to 300 kDamolecular weight cut off (MWCO) centrifugation at4,000 × g, 4 °C to concentrate, then washed twice with PBSto remove residual media components such as sugars. Theconcentrated samples were subsequently centrifuged at12,000 × g, 4 °C for 30 mins to eliminate microvesiclesbefore pelleting the remaining exosomes via ultra-centrifugation at 100,000 × g, 4 °C on a 5.5% sucrose padfor 15 h. Isolated exosomes were re-suspended in PBS andquantified prior to further analysis. Isolated exosomes werediluted in PBS and analyzed using the Nanosight NS300System (Malvern Instruments, UK), which is equipped witha blue laser (405 nm). The Nanosight Tracking Analysissoftware was used to provide particle concentrations andsize distribution profiles. For exosomal protein quantifica-tion, exosomes were mixed with equal volume of 10% SDSand boiled for 10 min to extract proteins from the membraneand inside the vesicles and then subsequently diluted 5-foldin PBS to dilute out SDS for BCA assay.

TMT labeling and liquid chromatography withtandem mass spectroscopy

A total of 250 μg denatured exosomes (generated either underNx or Hx conditions) were loaded into 15% SDS-PAGE gelsfor electrophoresis at 100 V for 30min. Each gel band wascut into small pieces and washed with 25mM triethy-lammonium bicarbonate (TEAB) in 50% acetonitrile (ACN).The gel pieces were then dehydrated with 100%ACN andvacuum dried. Reduction was carried out using 5mM Tris 2-carboxyethyl phosphine hydrochloride (TCEP) in 50mMTEAB buffer at 60 °C for 60min, followed by alkylation with55 mM iodoacetamide (IAA) in 50mM TEAB buffer at roomtemperature for 30min. TCEP and IAA were removed byperforming alternate washes with 50mM TEAB buffer and50mM TEAB in 50% ACN. The gel pieces were thendehydrated and dried for a second time. The gel pieces werecompletely rehydrated with 10 ng/ml of sequencing-grademodified trypsin solution (Promega Corporation, Madison,WI, USA) in 100mM TEAB (protein to trypsin ratio of 1:50)on ice and subsequently incubated at 37 °C for overnightdigestion. Tryptic peptides were extracted using 50% ACNand 5% acetic acid before being dried by vacuum cen-trifugation (Eppendorf, USA). Labeling was performed usingTMT reagent multiplex kits according to the manufacturer’sprotocol (Thermo Scientific, USA). The TMT-labeled

Hypoxia-induced tumor exosomes promote M2-like macrophage polarization of infiltrating myeloid cells. . .

peptides were pooled and desalted using Sep-Pak C18 Vaccartridges (Waters, Milford, MA) before being vacuum-centrifuged to dry. Details of the labeling scheme used areprovided in the Supplementary Data. The labeled sampleswere combined prior to fractionation on a Xbridge™ C18column (4.6 × 250mm, Waters, Milford, MA, USA) forsubsequent analysis by LC-MS/MS.

The fractionated peptides were separated and analyzedusing a Dionex Ultimate 3000 RSLCnano system coupledto a Q Exactive instrument (Thermo Fisher Scientific, MA,USA) as previously described [53]. The raw data wereconverted to Mascot generic file format using Proteome-Discoverer™ v1.4 software (PD, Thermo Scientific, SanJose, USA). The MS/MS spectra were deisotoped anddeconvoluted using the MS2 spectrum processor node inPD. Protein identification and quantitation were performedby comparing MS/MS spectra against the Uniprot Humandatabase (released on 7/25/2016, 70,849 sequences,23,964,784 residues) using Mascot software version 2.41(Matrix Sciences, London, UK). A threshold of maximumtwo missed trypsin cleavages was applied. Peptide pre-cursor mass tolerances of 10 ppm and 0.02 Da were usedduring data search. TMT6 quantitation was used to measurethe relative protein expression level. Fixed modificationswere carbamidomethylation (+57.021 Da) of cysteine resi-dues. Variable modifications were deamidation (+0.984Da) of asparagine and glutamine residues, as well as oxi-dation (+15.995 Da) of methionine residues. All statisticalanalysis was performed using experimental triplicate.Reproducibility between the replicates was analyzed usinglinear regression analysis method. Statistical significancewas determined by Student's T test. A frequency distributionanalysis was used to determine the fold-change cutoff forrelative protein abundance. Significantly changed proteinswere determined by the volcano plot and used for thesubsequent analysis (Supplementary Fig. S3).

Bone marrow cells isolation, differentiation, andexosome treatment

Eight-week-old male, B57CL/6NTac mice (InVivos, Pte.Ltd., Singapore) were used in all in vivo experiments. Bonemarrow cells were extracted from mouse femurs and tibiasvia flushing with RPMI media through a 25 G needle. Thesamples were then passed through 40-µm cell strainers toobtain single cell suspension. Next, the cells were cen-trifuged at 300 × g, 25 °C for 10 min, re-suspended in redblood cell lysis buffer, and incubated for 10 min with gentleagitation every 60 s. The resultant cell suspensions werethen pelleted via centrifugation and re-suspended inappropriate media/buffers as required for downstreamapplications. To obtain BMMs, isolated bone marrow cellswere cultured for 7 days in RPMI medium supplemented

with 10% FBS and M-CSF (25 ng/ml, BioLegend, SanDiego, CA, USA). For generation of BMDCs, isolated bonemarrow cells were cultured for 7 days in RPMI mediumsupplemented with 10% FBS and GM-CSF (20 ng/ml,BioLegend).

To examine the exosome-mediated macrophage polar-ization, 5 µg/ml of normoxic or hypoxic exosomes weretreated on BMMs supplemented with 10% exosome-freeFBS (Gibco exosome-depleted FBS, Thermo Fisher Sci-entific) and M-CSF (25 ng/ml) for 2 days. IL-4 (20 ng/ml,BioLegend) was used for M2-type macrophage polarization(Supplementary Fig. S5). Exosome-treated cells were sub-jected to western blot analysis, RT-qPCR, and FACSanalysis.

THP-1 cell differentiation and exosome treatment

Differentiation of THP-1 monocytic cells into amacrophage-like (M0) phenotype was done by treatmentwith 20 ng/ml phorbol-12-myristate-13-acetate (PMA;Sigma-Aldrich) for 24 h in 5% FBS–RPMI media followedby resting in 10% FBS–RPMI control media for 24 h.Following treatment, differentiated cells were washed twicewith PBS before being dissociated with cell dissociationsolution (Invitrogen) and seeded onto 5-cm dish and 5 µg/ml of exosomes in 10% exosome-free FBS–RPMI mediawere treated for 48 h.

Macrophage invasion assay

Mouse RAW 264.7 macrophages were seeded onto Trans-well® inserts coated with Matrigel and cultured in a 24-wellplate with 10 µg of normoxic or hypoxic exosome-containing medium (1% exosome-free FBS-DMEM) inlower chamber. Cell invasion in response to normoxic orhypoxic exosomes was monitored for 24 h and migratedcells on the bottom membrane were fixed, visualized, andquantitated by crystal violet staining.

Western blot analysis

BMMs were treated with exosomes as described above.Cells were washed in ice-cold PBS and lysed in modifiedRIPA buffer (50 mM Tris–HCl, 150 mM NaCl, 1% NP-40,pH 8.0, 1× protease inhibitor cocktail, phosphatase inhibi-tors). Lysates were clarified by centrifugation (16,000 × g,30 min) and subjected to western blotting using the indi-cated primary antibodies at 1:1000 dilution. Protein–anti-body conjugates were visualized using achemiluminescence detection kit (Thermo Fisher Scien-tific). Antibodies against exosomal markers, Alix, CD63,and TSG101, were obtained from Santa Cruz Biotechnol-ogy (Santa Cruz, USA). Antibodies to HIF-1α, STAT1,

J. E. Park et al.

pSTAT1 Y701, STAT6, pSTAT6 Y641, mTOR, p-mTORS2448, pS6K T398, p4E-BP T37/46, AKT, and pAKTS473 were purchased from Cell Signaling Technologies(Danvers, MA, USA). Actin, tubulin, and GAPDH anti-bodies were obtained from Millipore (Billerica, MA).

Total RNA and miRNA extraction and RT-qPCR

BMMs were treated with exosomes as described above andsubjected to total RNA or miRNA extraction. Total RNAextraction was performed using Nucleospin RNA kits(MACHEREY-NAGEL GmbH & Co.) according to themanufacturer’s protocol. RT-qPCR was performed using aCFX96 Real-Time PCR Detection System (Bio-Rad) withKAPA SYBR® FAST qPCR Master Mix. Actin or 18 sRNA were used as internal controls. To detect the level ofmiR-let-7a, we used mirVanaTM miRNA isolation kit(Invitrogen) to isolate miRNA according to the manu-facturer’s protocol and subsequently, miRNA was reversetranscribed using miRCURY LNATM Universal cDNASynthesis Kit (Exiqon, Denmark). Synthetic UniSp6, aspike-in control, was used to monitor the efficiency of theRT reaction. The primers for U6 and miR-let-7a wereobtained from Exiqon.

We transfected 50 nM of mse-let-7a-mimics (UGAG-GUAGUAGGUUGUAUAG) to mouse Raw264.7 cellsusing jetPEI-macrophage transfection reagent (PolyPlusTransfection, Illkirch, France) and tested for expression ofinsulin signaling molecules by RT-qPCR after 24 h. Theprimer sequences used for real-time PCR are provided in theSupplementary Information (Supplementary Table S5).

Leukocyte infiltration assay

Eight-week-old male, B57CL/6NTac mice were purchasedfrom InVivos and randomly assigned to different treatinggroups for leukocyte infiltration assay. A total of 1 × 106

B16-F0 tumor cells in growth factor-reduced Matrigel werepremixed with either normoxic or hypoxic exosomes (50µg) and then administered via s.c. injection into the back ofB57CL/6NTac mice (n= 6 per group). Tumor sizes werethen assessed daily by routine caliper measurement. Micewere euthanized when the tumor size exceeded 1.5 cmdiameter, as per protocol guidelines. The tumors were thenexcised, finely minced, and enzymatically dissociated usingcollagenase D (2 mg/ml) and DNase I (20 µg/ml) for 30 minat 37 °C to obtain single cell suspensions. Isolated cellsuspension was subjected to antibody staining with anti-F4/80-FITC and anti-CD206-PE (Miltenyi Biotech GmbH,Bergisch Gladbach, Germany) for FACS analysis. All ani-mal studies were approved by an Institutional Animal Careand Use Committee (IACUC, ARF-SBS/NIE-A0286) andwere performed in accordance with approved guidelines and

regulations of the Animal Facility Center of the School ofBiological Sciences, Nanyang Technological University,Singapore.

Flow cytometric analysis for exosome-mediatedBMM polarization

We used three different batches of exosome samples toexamine the effects and reproducibility of exosome treat-ment on BMM polarization. Normoxic or hypoxic exo-somes were treated as described above and further stainedwith Fluorochrome-conjugated antibodies, CD11c-APC,MHCII-PerCP-Vio700TM, F4/80-FITC, CD3-PerCP-Vio®700, CD25-PE, CD8a-APC, CD4-FITC, and CD135-PerCP-Vio700TM (Miltenyi Biotech, Bergisch Gladbach,Germany) and CD206-PE (Biolegend). M2-type polariza-tion of THP-1 cells was detected with CD163-PE (Biole-gend). Labeled cells were analyzed using a FACSCaliburflow cytometer (BD, San Jose, CA) and FlowJo software(Tree Star Inc., Oregon, USA).

Metabolic assays

OCR of BMMs was measured with a XF24 extracellularflux analyzer (Seahorse Bioscience). Briefly, BMMsunderwent the same exosome treatment for polarization and4 × 104 exosome-treated BMMs cells/well were seeded in aXF-24 cell culture microplate, a part of Seahorse XF24FluxPak (Seahorse Bioscience, Agilent Technologies, SantaClara, CA, USA), and OCR measurements were normalizedto cell number. Cells were initially plated in XF Seahorsemedia with both glucose and glutamine in mitochondrialstress test using the following concentrations of injectedcompounds, as according to the manufacturer’s standardprotocol (XF cell mito stress test kit, Seahorse Bioscience):OM, 1 μM; ROT, 0.75 μM; electron transport chain accel-erator p-trifluoromethoxy carbonyl cyanide phenyl hydra-zine (FCCP), 0.5 μM; AA, 1.5 μM. OCR test wasindependently performed with three different batches ofexosome samples to test the effect of exosomes on meta-bolic changes.

Statistical analysis

Statistical analyses were performed using SPSS18 software(v18.0; SPSS, Chicago, IL, USA). Differences betweengroups were assessed by Student’s t-test and p < 0.05 wasconsidered as significant.

Acknowledgements This work is in part supported by grants from theSingapore Ministry of Education (MOE2014-T2-2-043, MOE2016-T2-2-018, and MOE2016-T3-1-003) and the National MedicalResearch Council of Singapore (NMRC-OF-IRG-0003-2016).

Hypoxia-induced tumor exosomes promote M2-like macrophage polarization of infiltrating myeloid cells. . .

Author contributions JEP designed and performed the experiments,analyzed the data, and wrote the paper; BD performed the exosomeTMT proteomics experiments and analyzed the data; SWT and NGperformed animal experiments; CFT performed exosome preparationand analysis; and JKL, KWY, OLK, and JPT contributed to reagentsand discussion; SSK conceived, designed, supervised the project, andrevised the manuscript. All co-authors contributed to the revision ofthe manuscript.

Compliance with ethical standards

Conflict of interest The authors declare that they have no conflict ofinterest.

Publisher’s note: Springer Nature remains neutral with regard tojurisdictional claims in published maps and institutional affiliations.

References

1. Park JE, Tan HS, Datta A, Lai RC, Zhang H, Meng W, et al.Hypoxic tumor cell modulates its microenvironment to enhanceangiogenic and metastatic potential by secretion of proteins andexosomes. Mol Cell Proteom. 2010;9:1085–99.

2. Wilson WR, Hay MP. Targeting hypoxia in cancer therapy. NatRev Cancer. 2011;11:393–410.

3. Colombo M, Raposo G, Théry C. Biogenesis, secretion, andintercellular interactions of exosomes and other extracellularvesicles. Annu Rev Cell Dev Biol. 2014;30:255–89.

4. De Toro J, Herschlik L, Waldner C, Mongini C. Emerging roles ofexosomes in normal and pathological conditions: new insights fordiagnosis and therapeutic applications. Front Immunol.2015;4:203.

5. Simons M, Raposo G. Exosomes—vesicular carriers for inter-cellular communication. Curr Opin Cell Biol. 2009;21:575–81.

6. Liu Y, Gu Y, Han Y, Zhang Q, Jiang Z, Zhang X, et al. Tumorexosomal rnas promote lung pre-metastatic niche formation byactivating alveolar epithelial TLR3 to recruit neutrophils. CancerCell. 2016;30:243–56.

7. Peinado H, Alečković M, Lavotshkin S, Matei I, Costa-Silva B,Moreno-Bueno G, et al. Melanoma exosomes educate bone mar-row progenitor cells toward a pro-metastatic phenotype throughMET. Nat Med. 2012;18:883–91.

8. Wu L, Zhang X, Zhang B, Shi H, Yuan X, Sun Y, et al. Exosomesderived from gastric cancer cells activate NF-κB pathway inmacrophages to promote cancer progression. Tumor Biol.2016;37:12169–80.

9. Boelens Mirjam C, Wu Tony J, Nabet Barzin Y, Xu B, Qiu Y,Yoon T, et al. Exosome transfer from stromal to breast cancercells regulates therapy resistance pathways. Cell. 2014;159:499–513.

10. Hoffman RM. Stromal-cell and cancer-cell exosomes leading themetastatic exodus for the promised niche. Breast Cancer Res.2013;15:310.

11. Robbins PD, Morelli AE. Regulation of immune responses byextracellular vesicles. Nat Rev Immunol. 2014;14:195–208.

12. Viaud S, Terme M, Flament C, Taieb J, André F, Novault S, et al.Dendritic cell-derived exosomes promote natural killer cell acti-vation and proliferation: a role for NKG2D ligands and IL-15Rα.PLoS ONE. 2009;4:e4942.

13. Bretz NP, Ridinger J, Rupp A-K, Rimbach K, Keller S, Rupp C,et al. Body fluid exosomes promote secretion of inflammatorycytokines in monocytic cells via toll-like receptor signaling. J BiolChem. 2013;288:36691–702.

14. Whiteside TL. Exosomes and tumor-mediated immune suppres-sion. J Clin Invest. 2016;126:1216–23.

15. Espinoza JL, Takami A, Yoshioka K, Nakata K, Sato T, KasaharaY, et al. Human microRNA-1245 down-regulates the NKG2Dreceptor in natural killer cells and impairs NKG2D-mediatedfunctions. Haematologica. 2012;97:1295–303.

16. Zhou J, Wang S, Sun K, Chng W-J. The emerging roles of exo-somes in leukemogeneis. Oncotarget. 2016;7:50698–707.

17. Lobb RJ, Lima LG, Möller A. Exosomes: key mediators ofmetastasis and pre-metastatic niche formation. Semin Cell DevBiol. 2017;67:3–10.

18. Kucharzewska P, Belting M. Emerging roles of extracellularvesicles in the adaptive response of tumour cells to micro-environmental stress. J Extracell Vesicles. 2013;2:1–10.

19. Villarroya-Beltri C, Baixauli F, Gutierrez-Vazquez C, Sanchez-Madrid F, Mittelbrunn M. Sorting it out: regulation of exosomeloading. Semin Cancer Biol. 2014;28:3–13.

20. Wargo JA, Reddy SM, Reuben A, Sharma P. Monitoring immuneresponses in the tumor microenvironment. Curr Opin Immunol.2016;41:23–31.

21. Storm MP, Sorrell I, Shipley R, Regan S, Luetchford KA, SathishJ, et al. Hollow fiber bioreactors for in vivo-like mammalian tissueculture. J Vis Exp. 2016:53431.

22. Thompson A, Schäfer J, Kuhn K, Kienle S, Schwarz J, SchmidtG, et al. Tandem mass tags: a novel quantification strategy forcomparative analysis of complex protein mixtures by MS/MS.Anal Chem. 2003;75:1895–904.

23. Semenza GL. HIF-1 and tumor progression: pathophysiology andtherapeutics. Trends Mol Med. 2002;8:S62–7.

24. Zhu Q, Wong AK, Krishnan A, Aure MR, Tadych A, Zhang R,et al. Targeted exploration and analysis of large cross-platformhuman transcriptomic compendia. Nat Methods. 2015;12:211–4.

25. Lin EY, Nguyen AV, Russell RG, Pollard JW. Colony-stimulatingfactor 1 promotes progression of mammary tumors to malignancy.J Exp Med. 2001;193:727–40.

26. Qian B-Z, Li J, Zhang H, Kitamura T, Zhang J, Campion LR,et al. CCL2 recruits inflammatory monocytes to facilitate breasttumor metastasis. Nature. 2011;475:222–5.

27. Gray CP, Arosio P, Hersey P. Association of increased levels ofheavy-chain ferritin with increased CD4+CD25+regulatory T-cell levels in patients with melanoma. Clin Cancer Res.2003;9:2551–9.

28. Marie JC, Letterio JJ, Gavin M, Rudensky AY. TGF-β1 maintainssuppressor function and Foxp3 expression in CD4+CD25+reg-ulatory T cells. J Exp Med. 2005;201:1061–7.

29. Kumar V, Patel S, Tcyganov E, Gabrilovich DI. The nature ofmyeloid-derived suppressor cells in the tumor microenvironment.Trends Immunol. 2016;37:208–20.

30. Ugel S, De Sanctis F, Mandruzzato S, Bronte V. Tumor-inducedmyeloid deviation: when myeloid-derived suppressor cells meettumor-associated macrophages. J Clin Invest. 2015;125:3365–76.

31. Green CE, Liu T, Montel V, Hsiao G, Lester RD, Subramaniam S,et al. Chemoattractant signaling between tumor cells and macro-phages regulates cancer cell migration, metastasis and neovascu-larization. PLoS ONE. 2009;4:e6713.

32. Damuzzo V, Pinton L, Desantis G, Solito S, Marigo I, Bronte V,et al. Complexity and challenges in defining myeloid-derivedsuppressor cells. Cytometry B Clin Cytom. 2015;88:77–91.

33. Kikushige Y, Yoshimoto G, Miyamoto T, Iino T, Mori Y, IwasakiH, et al. Human Flt3 is expressed at the hematopoietic stem celland the granulocyte/macrophage progenitor stages to maintain cellsurvival. J Immunol. 2008;180:7358–67.

34. Geeraerts X, Bolli E, Fendt S-M, Van Ginderachter JA. Macro-phage metabolism as therapeutic target for cancer, atherosclerosis,and obesity. Front Immunol. 2017;8:289.

J. E. Park et al.

35. Mosser DM, Edwards JP. Exploring the full spectrum of macro-phage activation. Nat Rev Immunol. 2008;8:958–69.

36. Roszer T. Understanding the mysterious M2 macrophage throughactivation markers and effector mechanisms. Mediators Inflamm.2015;2015:816460.

37. Genin M, Clement F, Fattaccioli A, Raes M, Michiels C. M1 andM2 macrophages derived from THP-1 cells differentially mod-ulate the response of cancer cells to etoposide. BMC Cancer.2015;15:577–577.

38. Hu JM, Liu K, Liu JH, Jiang XL, Wang XL, Chen YZ, et al.CD163 as a marker of M2 macrophage, contribute to predicteaggressiveness and prognosis of Kazakh esophageal squamouscell carcinoma. Oncotarget. 2017;8:21526–38.

39. Sudan B, Wacker MA, Wilson ME, Graff JW. A systematicapproach to identify markers of distinctly activated human mac-rophages. Front Immunol. 2015;6:253.

40. Biswas Subhra K. Metabolic reprogramming of immune cells incancer progression. Immunity. 2015;43:435–49.

41. Otero-Albiol D, Felipe-Abrio B. MicroRNA regulating metabolicreprogramming in tumor cells: new tumor markers. Cancer TranslMed. 2016;2:175–81.

42. Zhu H, Shyh-Chang N, Segrè Ayellet V, Shinoda G, Shah SamarP, Einhorn William S, et al. The Lin28/let-7 axis regulates glucosemetabolism. Cell. 2011;147:81–94.

43. Lewis BP, Burge CB, Bartel DP. Conserved seed pairing, oftenflanked by adenosines, indicates that thousands of human genesare microRNA targets. Cell. 2004;120:15–20.

44. Finger EC, Giaccia AJ. Hypoxia, inflammation, and the tumormicroenvironment in metastatic disease. Cancer Metastasis Rev.2010;29:285–93.

45. Kao J, Houck K, Fan Y, Haehnel I, Libutti SK, Kayton ML, et al.Characterization of a novel tumor-derived cytokine. Endothelial-monocyte activating polypeptide II. J Biol Chem.1994;269:25106–19.

46. Kore RA, Edmondson JL, Jenkins SV, Jamshidi-Parsian A, DingsRPM, Reyna NS, et al. Hypoxia-derived exosomes induce puta-tive altered pathways in biosynthesis and ion regulatory channelsin glioblastoma cells. Biochem Biophys Rep. 2018;14:104–13.

47. Elliott LA, Doherty GA, Sheahan K, Ryan EJ. Human tumor-infiltrating myeloid cells: phenotypic and functional diversity.Front Immunol. 2017;8:86.

48. Galván-Peña S, O’Neill LAJ. Metabolic reprograming in macro-phage polarization. Front Immunol. 2014;5:420.

49. O’Neill LAJ, Pearce EJ. Immunometabolism governs dendriticcell and macrophage function. J Exp Med. 2016;213:15–23.

50. Wenes M, Shang M, Di Matteo M, Goveia J, Martín-Pérez R,Serneels J, et al. Macrophage metabolism controls tumor bloodvessel morphogenesis and metastasis. Cell Metab. 2016;24:701–15.

51. Jérôme T, Laurie P, Louis B, Pierre C. Enjoy the silence: the storyof let-7 microRNA and cancer. Curr Genomics. 2007;8:229–33.

52. Young L, Sung J, Stacey G, Masters JR. Detection of mycoplasmain cell cultures. Nat Protoc. 2010;5:929.

53. Park JE, Sun Y, Lim SK, Tam JP, Dekker M, Chen H, et al.Dietary phytochemical PEITC restricts tumor development viamodulation of epigenetic writers and erasers. Sci Rep.2017;7:40569.

54. Hume DA, MacDonald KPA. Therapeutic applications of mac-rophage colony-stimulating factor-1 (CSF-1) and antagonists ofCSF-1 receptor (CSF-1R) signaling. Blood. 2012;119:1810–20.

55. Murdoch C, Giannoudis A, Lewis CE. Mechanisms regulating therecruitment of macrophages into hypoxic areas of tumors andother ischemic tissues. Blood. 2004;104:2224–34.

56. Nielsen SR, Schmid MC. Macrophages as key drivers ofcancer progression and metastasis. Mediators Inflamm.2017;2017:11.

57. Li M, Knight DA, A Snyder L, Smyth MJ, Stewart TJ. A role forCCL2 in both tumor progression and immunosurveillance.Oncoimmunology. 2013;2:e25474.

58. Lee DD, Lal CV, Persad EA, Lowe C-W, Schwarz AM, AwasthiN, et al. Endothelial monocyte-activating polypeptide II mediatesmacrophage migration in the development of hyperoxia-inducedlung disease of prematurity. Am J Respir Cell Mol Biol.2016;55:602–12.

59. Jia W, Kidoya H, Yamakawa D, Naito H, Takakura N. Galectin-3accelerates M2 macrophage infiltration and angiogenesis intumors. Am J Pathol. 2013;182:1821–31.

60. Alkhateeb AA, Han B, Connor JR. Ferritin stimulates breastcancer cells through an iron-independent mechanism and islocalized within tumor-associated macrophages. Breast CancerRes Treat. 2013;137:733–44.

61. Gray CP, Arosio P, Hersey P. Heavy chain ferritin activatesregulatory T cells by induction of changes in dendritic cells.Blood. 2002;99:3326–34.

62. Alkhateeb AA, Connor JR. The significance of ferritin in cancer:anti-oxidation, inflammation and tumorigenesis. Biochim BiophysActa. 2013;1836:245–54.

63. Jezequel P, Campion L, Spyratos F, Loussouarn D, Campone M,Guerin-Charbonnel C, et al. Validation of tumor-associated mac-rophage ferritin light chain as a prognostic biomarker in node-negative breast cancer tumors: a multicentric 2004 national PHRCstudy. Int J Cancer. 2012;131:426–37.

64. Wu T, Li Y, Liu B, Zhang S, Wu L, Zhu X, et al. Expression offerritin light chain (FTL) Is elevated in glioblastoma, and FTLsilencing inhibits glioblastoma cell proliferation via the GADD45/JNK pathway. PLoS ONE. 2016;11:e0149361.

65. Bellomo C, Caja L, Moustakas A. Transforming growth factor βas regulator of cancer stemness and metastasis. Br J Cancer.2016;115:761.

66. Lebrun J-J. The dual role of TGF in human cancer: from tumorsuppression to cancer metastasis. ISRN Mol Biol. 2012;2012:28.

67. Tauriello DVF, Palomo-Ponce S, Stork D, Berenguer-Llergo A,Badia-Ramentol J, Iglesias M, et al. TGFβ drives immune evasionin genetically reconstituted colon cancer metastasis. Nature.2018;554:538.

68. Chen X, Wang S, Wu N, Yang CS. Leukotriene A4 hydrolase as atarget for cancer prevention and therapy. Curr Cancer Drug Tar-gets. 2004;4:267–83.

69. Vo TTL, Jang WJ, Jeong CH. Leukotriene A4 hydrolase: anemerging target of natural products for cancer chemopreventionand chemotherapy. Ann N Y Acad Sci. 2018;1431:3–13.

70. Balogh KN, Templeton DJ, Cross JV. Macrophage MigrationInhibitory Factor protects cancer cells from immunogenic celldeath and impairs anti-tumor immune responses. PLoS ONE.2018;13:e0197702.

71. Kindt N, Journe F, Laurent G, Saussez S. Involvement of mac-rophage migration inhibitory factor in cancer and novel ther-apeutic targets. Oncol Lett. 2016;12:2247–53.

72. Chen SC, Kung ML, Hu TH, Chen HY, Wu JC, Kuo HM, et al.Hepatoma-derived growth factor regulates breast cancer cellinvasion by modulating epithelial--mesenchymal transition. JPathol. 2012;228:158–69.

73. Gialeli C, Theocharis AD, Karamanos NK. Roles of matrixmetalloproteinases in cancer progression and their pharmacolo-gical targeting. FEBS J. 2011;278:16–27.

Hypoxia-induced tumor exosomes promote M2-like macrophage polarization of infiltrating myeloid cells. . .

74. Wang D, Zhang S, Chen F. High expression of PLOD1 drivestumorigenesis and affects clinical outcome in gastrointestinalcarcinoma. Genet Test Mol Biomarkers. 2018;22:366–73.

75. Gilkes DM, Bajpai S, Wong CC, Chaturvedi P, Hubbi ME, Wirtz D,et al. Procollagen lysyl hydroxylase 2 is essential for hypoxia-induced breast cancer metastasis. Mol Cancer Res. 2013;11:456–66.

76. Mogami T, Yokota N, Asai-Sato M, Yamada R, Koizume S,Sakuma Y, et al. Annexin A4 is involved in proliferation, chemo-resistance and migration and invasion in ovarian clear cell ade-nocarcinoma cells. PLoS ONE. 2013;8:e80359.

77. Winter J, Diederichs S. Argonaute proteins regulate microRNAstability: Increased microRNA abundance by Argonaute

proteins is due to microRNA stabilization. RNA Biol.2011;8:1149–57.

78. Dueck A, Ziegler C, Eichner A, Berezikov E, Meister G. micro-RNAs associated with the different human Argonaute proteins.Nucleic Acids Res. 2012;40:9850–62.

79. Leca J, Martinez S, Lac S, Nigri J, Secq V, Rubis M, et al. Cancer-associated fibroblast-derived annexin A6+extracellular vesiclessupport pancreatic cancer aggressiveness. J Clin Invest.2016;126:4140–56.

80. Qi H, Liu S, Guo C, Wang J, Greenaway FT, Sun M-Z. Role ofannexin A6 in cancer. Oncol Lett. 2015;10:1947–52.

J. E. Park et al.